Which is why it is such a pity that they are still putting so much valuable work into fighting Elisp instead of having a full fledged CL with a native compiler under the hood. The FSF even has two own, GPL'ed CL implementations, CLISP and GCL, but because of RMS' hatred of CL, let them bitrot and never really cared to use them for anything.
I'm not contesting that, just saying that it is a pity (from the perspective of the wealth of the overall Lisp ecosystem) that all this valuable work has been done in Elisp and not in Common Lisp. Moving to Guile will probably take several years, and even if it really works some day, it will be just another insular Emacs-only solution, because nobody else is using Guile. It wont benefit the overall Lisp ecosystem in any way.
That way authors could transition from elisp to common lisp.
Is CL really that much better than Scheme?
From a pragmatic standpoint, all standard-conforming Common Lisp code runs on a conformant implementation. Schemes often tend to diverge (and they have 6 different Rn standards). So I think CL is 'better' than Scheme, for that value of better.
Correct or not, I've tended to think of Emacs as an LVM ([e]Lisp Virtual Machine) for years. The "bloated kitchen sink" argument holds little sway with me any more than saying that OpenBSD is a bloated kitchen sink because it can run a lot of software too. Emacs happens to be an awesome Batteries Included LVM.
This is not an accident. GNU Emacs was probably intended to be the first step towards a complete lisp system for GNU, akin to zmacs on symbolics.
The free software manifesto mentions lisp several times http://www.gnu.org/gnu/manifesto.html
Each new Emacs version shows little on the surface which you might expect from a software suite that has had everything and the kitchen sink since the 90's, but in the inside there's an ongoing morphosis into something greater, that quietly reveals itself version by version if only you take the time to study not only what's new but how the new things work. Then suddenly, say, five years later, a new usage paradigm emerges seemingly out of nothing eventhough Emacs has had the bits and pieces in for many revisions already. This is just fascinating. For example, the various completion subroutines have slowly converged into a more solid, generic completion subsystem. And best of all, you have the source code to play with.
If Emacs is nearly an operating system then it's kind of nearly a software Lisp machine.